AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

CLadder: Assessing Causal Reasoning in Language Models

Zhijing Jin, Yuen Chen, Felix Leeb, Luigi Gresele, Ojasv Kamal, Zhiheng LYU, Kevin Blin, Fernando Gonzalez Adauto, Max Kleiman-Weiner, Mrinmaya Sachan, Bernhard Schölkopf

NeurIPS 2023 · Large language models › Language, safety and interpretability · approach: Text generation and evaluation metrics · 16 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Apply six-step causal-inference chain-of-thought prompting instead of direct promptingGPT-4 · CLadder v1.5accuracy: 62.03 -> 70.40; 8.37 pointsImproved
Apply six-step causal-inference chain-of-thought prompting instead of direct promptingGPT-4 · CLadder v1.5, rung 1accuracy: 63.01 -> 83.35Improved
Apply six-step causal-inference chain-of-thought prompting instead of direct promptingGPT-4 · CLadder v1.5, rung 2accuracy: 62.82 -> 67.47Improved
Apply six-step causal-inference chain-of-thought prompting instead of direct promptingGPT-4 · CLadder v1.5, rung 3accuracy: 60.55 -> 62.05Improved
Apply six-step causal-inference chain-of-thought prompting instead of direct promptingGPT-4 · CLadder v1.0accuracy: 64.28 -> 66.64; 2.36 pointsImproved
Use CLadder CoT on anti-commonsensical examples rather than vanilla GPT-4GPT-4 · CLadder v1.5, anti-commonsensicalaccuracy: 60.47 -> 70.12; 9.65 pointsImproved
Remove Step 1, causal graph extraction (ablation)CLadder CoT · CLadder v1.0accuracy: 66.64 -> 64.54Hurt
Remove Step 2, query-type identification (ablation)CLadder CoT · CLadder v1.0accuracy: 66.64 -> 63.74Hurt
Remove Step 3, query formalization (ablation)CLadder CoT · CLadder v1.0accuracy: 66.64 -> 63.43Hurt
Remove Step 4, available-data extraction (ablation)CLadder CoT · CLadder v1.0accuracy: 66.64 -> 64.47Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy